COS 72-2 - How far back in time do we need to go? Modelling the dependence of current species to past land-use and climate changes

Abstract: The responses of communities to global changes are not necessarily instantaneous or synchronous, which means that communities are not in equilibrium with current environmental conditions. Time-lagged responses have already been quantified in several ways, arguing that current community composition and species distribution still depend on past anthropogenic pressures. However, it is still unknown how the effects of past land-use and climate changes decrease as we go back in time. Understanding this is particularly important for better informing future biodiversity changes. The goal of this study is to evaluate the effects of current and past climate and land-use conditions on the continental distribution of 92 European forest bird species, sampled during the period 2013-2017. Our approach to address this goal was first to use generalized linear regressions considering environmental predictors over 2013-2017. These null models assume an equilibrium of species distribution with current conditions (no time-lag effect). Second, using temporally-weighted generalized linear regressions, we quantified the temporal extent of the time-lag in explaining the current distribution of a given species in response to past land-use and climate changes (over 1950-2017). In these models, the effect size of each focal environmental predictor is associated to a weighting parameter that defines: (i) the temporal extent of the time lag that must be considered to capture the majority of the effects of past conditions on current species distribution, and (ii) the shape of an exponential negative curve that describes the temporal decrease of the effect size as we go back in time. Here we found that temporally-weighted generalized linear regressions are better (according to deviance tests) than the null models for the majority ( >85%) of forest bird species. This result confirms the prevalence of the non-equilibrium hypothesis. The temporal extent of the time lag varies among species and environmental variables and can reach up to several decades, consistent with former findings. The temporal extent of the time lag is slightly correlated to dispersal-related traits (such as body mass) suggesting which species have higher dependency to older environmental conditions. In conclusion, through this work, we introduce a new conceptual framework associated to an underlying modelling approach that quantify the temporal extent of the time-lagged response in the effect of a given environmental variable on current species distribution.